Reference no: EM133668897
Assignment for Big Data and Deep Learning
You have been given daily data on the following US technological stocks, Amazon, Intel and Facebook, from 13/02/2015 - 12/02/2020. Data were ob- tained from Yahoo Finance. There is one .csv file for every stock available in Keats for you to download.
Given the above dataset, students need to use a neural network of their own choice, to forecast whether the stock return of Amazon, will go up or down in the next day of trading, given data from the previous 14 days of trading.
As predictors, students need to use the volume and stock returns of all stocks. Further students need to create the predictive variable which will be 1 if the stock return is positive and 0 otherwise.
Your report should contain the following information:
Part A: In no more than two A4 pages (or 1,000 words), students need to discuss the main components of the Neural Network they are going to use and the reason behind their choice. Specifically, the Architecture of their choice, the activation function to be used and the Cross-Validation scheme to be employed. Students should also succinctly summarise how learning occurs in their selected neural network.
Part B: In a .py or .ipynb notebook please follow the following steps:
Import the necessary libraries.
Read the different datasets, using pandas.
Create the train-validation-test split. Specifically, use the dates given below:
train_start_date="2015-04-28" train_end_date="2017-12-31" val_start_date="2018-01-03" val_end_date="2018-12-31" test_start_date="2019-01-02" test_end_date="2020-01-31"
Get the returns for every stock and standardise both the returns and the volume.
Create the forecasting variable, i.e. what we called label in the tutorial. Create a new dataframe that contains all the predictors and the predictive variable.
Create the training, validation and testing generators using the above dates.
Create the neural network (in Keras) and choose the relevant hyper- parameters and
Follow the Cross-Validation scheme that you defined above, and forecast using the optimal parameters from CV. For this you can use the ran- dom search function from the tutorial but you can also use your own written one. The code should be well documented.
Comment on your findings.
Repeat the above steps, but now, use information only from the stock of
Amazon. Comment on your findings.
Append your code at the end of your report in an Appendix.